About
Before AI was part of the vocabulary, I was deep inside production systems — Linux environments, Chef and Ansible automation, Grafana dashboards, and incident response at companies like Facebook Reality Labs and Rackspace. 0+ years of keeping infrastructure alive builds a specific kind of thinking: you learn to read the system before you change the system.
That background — understanding how things break, how they recover, and how to communicate across technical boundaries — is the foundation everything else is built on.
AI changed what's possible for a solo builder. With Claude as a genuine collaborator — not autocomplete — I can now ship things that used to require a full team: a manuscript editor with chapter-level continuity tracking, a force-directed graph visualization with sentiment analysis, a form system with embedded AI processing.
The work isn't about using AI instead of thinking. It's about using deep systems knowledge to ask the right questions, then using AI to build the answers faster than was previously possible.
How I Work
Every problem lives inside a larger system. Before writing a line of code, I map the structure — data flows, failure modes, user needs, edge cases. Fifteen years of incident response trained this instinct.
Claude isn't a shortcut — it's a collaborator. I use it for architecture review, debugging subtle logic, generating test cases, and iterating on design decisions at a speed that would be impossible alone.
Sysadmin discipline means caring about what happens after deployment. Documentation that people actually read. Monitoring that catches problems before users do. Code that the next person can understand.
Projects
AI developmental editor for self-published authors. Analyzes manuscripts chapter-by-chapter, generates story bibles from opening chapters, and flags continuity errors — mimicking how a human editor builds context over time. Built on Anthropic's Claude API.
View on GitHubTransforms Reddit comment threads into interactive force-directed graphs. Nodes sized by upvote score, colored by AFINN sentiment analysis. Pan, zoom, filter by depth and score. Makes large conversation topology immediately visible.
View on GitHubAI-powered intake form system that reduces manual data entry through intelligent form processing. Built with Next.js and TypeScript, deployable as an embeddable component. Active development with open pull requests.
View on GitHubTestimonials
Contact
Open to interesting problems.